K-Anonymization or why Aggregation is not Enough